Asana cuts model costs 76x in browser tests with GPT-6.1 Sol

Asana reported that by leveraging a GPT-6 model within Codex, its browser-based AI agent achieved a 76-fold reduction in per-query costs and a 5x improvement in response speed during internal testing. The company framed the optimization as a step toward offering customers access to more capable models without prohibitive expense.

Browser agents are an increasingly important category in enterprise AI, as they automate repetitive web-based tasks such as data entry, form filling, and workflow orchestration. Cost efficiency is a critical bottleneck for scaling such agents to broad user bases, since each interaction can involve multiple model calls.

The 76x cost reduction suggests significant gains from model-level optimization rather than architectural changes alone. If these results hold in production, Asana could deploy higher-tier models to a wider customer segment, potentially improving task automation quality while keeping subscription pricing stable.

This case also illustrates a broader trend in the AI industry: enterprises are increasingly focused on inference economics, not just model capability. As AI agents move from demos to daily-use tools, the ability to run them affordably at scale will be a key competitive differentiator.

This page provides an editorial summary based on publicly available information. It is not a republished article. Use the source link below for the original report.